This paper reviews existing methods for evaluating image segmentation. Segmentation evaluation methods are classified into three groups: analytical, empirical goodness, and empirical discrepancy. Analytical methods assess segmentation algorithms by analyzing their principles and properties, while empirical methods evaluate algorithms by comparing segmentation results with reference images. Empirical methods are further divided into goodness methods, which measure the quality of segmented images, and discrepancy methods, which compare segmented images with reference images to assess performance. The paper discusses the characteristics, advantages, and limitations of each method group. Analytical methods provide qualitative insights but are limited by the lack of general theory for image segmentation. Empirical goodness methods evaluate segmentation quality based on desired properties of segmented images, while empirical discrepancy methods compare segmented images with reference images to assess performance. Both types of empirical methods are widely used in segmentation evaluation. The paper also presents a comparative study of several empirical methods, including goodness-based and discrepancy-based methods. The study shows that discrepancy-based methods, such as those based on the probability of error and normalized distance, are more effective in evaluating segmentation performance. The results indicate that discrepancy-based methods provide more accurate and reliable evaluations compared to goodness-based methods. The paper concludes that segmentation evaluation is essential for improving the performance of existing segmentation algorithms and for developing new, more powerful algorithms. It emphasizes the need for further research and development in segmentation evaluation to enhance the accuracy and reliability of segmentation results.This paper reviews existing methods for evaluating image segmentation. Segmentation evaluation methods are classified into three groups: analytical, empirical goodness, and empirical discrepancy. Analytical methods assess segmentation algorithms by analyzing their principles and properties, while empirical methods evaluate algorithms by comparing segmentation results with reference images. Empirical methods are further divided into goodness methods, which measure the quality of segmented images, and discrepancy methods, which compare segmented images with reference images to assess performance. The paper discusses the characteristics, advantages, and limitations of each method group. Analytical methods provide qualitative insights but are limited by the lack of general theory for image segmentation. Empirical goodness methods evaluate segmentation quality based on desired properties of segmented images, while empirical discrepancy methods compare segmented images with reference images to assess performance. Both types of empirical methods are widely used in segmentation evaluation. The paper also presents a comparative study of several empirical methods, including goodness-based and discrepancy-based methods. The study shows that discrepancy-based methods, such as those based on the probability of error and normalized distance, are more effective in evaluating segmentation performance. The results indicate that discrepancy-based methods provide more accurate and reliable evaluations compared to goodness-based methods. The paper concludes that segmentation evaluation is essential for improving the performance of existing segmentation algorithms and for developing new, more powerful algorithms. It emphasizes the need for further research and development in segmentation evaluation to enhance the accuracy and reliability of segmentation results.